SAP Joule: Collaborative AI Agent powered by Open-Source LLMs

SAP announced today the expansion of its generative AI copilot Joule’s capabilities to support up to 80% of its customers’ most common business tasks. This will enable it to be a collaborative agent that can accomplish complex workflows.

For customers to get the most value from Joule and future innovations, they must be on SAP cloud platforms and systems. SAP is accelerating product innovation on the cloud in the hopes of attracting more customers to its RISE with SAP initiative.

Launched in Jan. 2021, RISE with SAP aims to guide customers’ transition from on-premises SAP ERP systems to the cloud, modernizing processes along the way. SAP reported that its cloud revenue increased by 25% in Q2 2024, with the Cloud ERP Suite growing by 33% as a result, demonstrating that RISE is effective.

Joule’s infusion of new features signals how serious SAP is about moving its customers to the cloud. SAP says on-premises customers can still use Joule. However, they will need to use the SAP Integration Suite to connect their existing infrastructure to SAP’s cloud services, enabling Joule to access and process data while extending AI capabilities to their on-premise environments.

Additional announcements at TechEd 2024 introduced additional open source large language model (LLM) support as part of the SAP Generative AI Hub, introduced the SAP Knowledge Graph, showcasing developer enhancements in SAP Build, highlighted specific use cases and reaffirmed the company’s commitment to upskill 2 million people by 2025.

SAP placing a strategic bet on Joule’s new agentic AI strengths  

Designed as a cloud-native AI assistant that is core to SAP’s Business Technology Platform (BTP), Joules’ ability to integrate and scale with all current and future apps, modules and platform environments to accelerate customers to the cloud further.

SAP made a strategic bet with BTP, believing their customers would see the value of a unified cloud platform over the legacy on-premises ERP systems, which earned a reputation for being challenging to integrate real-time data and third-party applications with. SAP doubling down on Joule shows they’re working to reverse their proprietary ways of the past and go after a more open cloud-based architecture that can deliver the accuracy, speed and scale their customers need.

 Embedded across SAP’s ecosystem, Joule can already understand business contexts, deliver data-driven insights and enable customers to get more work done using its advanced natural language processing and machine learning (ML) capabilities. With 80% of the most common business tasks now part of Joule, SAP is betting their latest gen AI copilot will be compelling enough for more customers to join RISE and move to the cloud.

SAP highlighted two use cases at TechEd 2024 to demonstrate the power of these agents:

• Dispute Management: AI agents autonomously resolve disputes related to invoices, credits and payments. This will significantly reduce manual intervention.
• Financial Accounting: Specialized agents streamline billing, ledger updates and invoice processing. This will ensure accuracy and efficiency.

“Collaborative AI agents from SAP represent a new era in enterprise productivity,” said Philipp Herzig, Chief Artificial Intelligence Officer at SAP. “Our ability to integrate multiple specialized AI agents into Joule allows businesses to automate intricate workflows and focus on tasks that truly require human ingenuity.”

Three new open-source models added to the SAP gen AI Hub

The open-source LLM announcements at TechEd 2024 show that SAP is continuing to develop its generative AI hub strategy. Open-source models now available on the SAP Generative AI Hub include Meta’s Llama 3.1 70B model, Mistral Large 2 (available by the end of 2024), and Mistral Codestral. SAP’s announcements this week at TechEd 2024 show that it is committed to keeping up with the quicker pace of innovation in open-source LLMs.

The SAP Generative AI Hub, positioned as a central node within SAP’s Business Technology Platform (BTP), connects both proprietary and open-source models. Developer tools like the Extensibility Wizard and SAP Build enhancements streamline this integration, reflecting SAP’s push toward a developer-friendly environment.

SAP continues to go on the offensive when it comes to providing more significant support for open-source LLMs. Their series of announcements this week at TechEd 2024 show they’re committed to keeping up with the quickening pace of innovation in open source LLMs in general and across all of open source strategically.

One of the main goals of going on the offensive with open-source LLMs is to enable enterprise-level standards for customers to adopt while ensuring reliability, scalability, security and performance, all within the SAP AI Core.

The following table summarizes the three open-source LLMs SAP announced support for during TechEd 2024:  

SAP’s new AI era has arrived

Long known for its dominance in the ERP market, SAP shows signs of successfully reinventing itself in a new AI era. Lessons learned on usability, the need for a more open, adaptive system architecture, and the need to provide customers with more flexibility in how they use data, including open-source LLMs, now dominate their product strategies. Their Business Technology Platform with an SAP AI Core reflects a more forward-thinking SAP that realizes their quickest path to value is recognizing customers need the freedom to go open source when they choose.